Dual E5649 PPD?

P

plext0r

[H]ard DCOTM x3
Joined
Dec 1, 2009
Messages
780
Has anyone successfully run bigadv on a dual E5649 system? I have one that is giving ECC errors even when I switch out the RAM. Not sure if I should reseat the CPUs and use some fresh thermal paste or what else could cause the known-good RAM to experience ECC errors. I went down to a single DIMM per CPU and only the second CPU's DIMM is complaining.

My dual X5650 system is getting around 130K PPD. The dual E5649 is claiming around 9K (frame time is 36:13). I imagine part of the problem is memory bandwidth since I went down to a single DIMM.

When I had this same system as my VMware vCenter server running 2008, it would reboot every couple of days. It doesn't reboot with Linux, but it's not behaving properly either.
 
In the past I ran e5649 on SR-2 doing bigadv. What MB are they on? Give us a little more info.
 
DooKey said:
In the past I ran e5649 on SR-2 doing bigadv. What MB are they on? Give us a little more info.
Click to expand...

Unfortunately these are in a commercial server-class machine, no SR-2. No ability to OC. I'm wondering if the hardware ECC errors are slowing the job way down or if this CPU is just too slow for bigadv.

All-core turbo appears to be 2.67GHz according to turbostat.
 
If the ram is known to be good, I would try disabling ECC support in the BIOS. One of the CPU's memory controller might be flaky.
 
2.66 is the bare minimum to make deadline on 8101 - you will be fine on all the others. Just to give you some idea my dual L5640 @ 2.4 does a frame in 36:54 to 38:32 dependant on WU.

You should make it with about 2 hours to spare
 
Nathan_P said:
2.66 is the bare minimum to make deadline on 8101 - you will be fine on all the others. Just to give you some idea my dual L5640 @ 2.4 does a frame in 36:54 to 38:32 dependant on WU.

You should make it with about 2 hours to spare
Click to expand...

Thanks for the info. I got the same 8101 WU frame time down to 33:03 by adding a couple of sticks of RAM (two channels per CPU). I'm hoping that when this WU is done, it will pull down a non-8101 bigadv WU so I can see how it performs.
 
An 8103 WU is running with a TPF of 25:53 and estimated PPD of 131K. That's more like it. The problem is the abundance of 8101 WUs where this guy won't complete on time and PPD suffers horribly.
 
You must log in or register to reply here.
Back
Top